[ppml] Reclamation (Was: Proposed Policy: PI assignments for V6)

[Geoff Huston]

Lee,

There are two sources of data that allow some estimates of this form to be 
made.

The first is the analysis of the IPv4 address space in terms of what is 
unused or not

One source of this data is at:
http://bgp.potaroo.net/ipv4/stats/allocated-all.html

Here I'll use the /8 notation meaning "16.7M addresses", rather than "a 
contiguous block of addresses mapped by a /8 address prefix".

143/8s are 'in play' right now (the other 93 /8s are either ietf reserved 
in various states or with IANA)

81/8s are advertised on the public global Internet.

62 /8s are 'unused' using this broad categorization

At a finer level 20 /8s are currently unallocated - in that there is no 
visible RIR information about its allocation. 10 /8s are in the historical 
address space (old B's mainly), which appear to be legacy unallocated 
addresses and the other 10 /8s are in the working pools of rir-managed 
address space (expansion windows, yet-to-be-allocated, etc)

The other 42 /8s are allocated, but have no visibility in the public routed 
Internet. They may or may not be in use in private contexts.

The other piece of data is a model about address consumption. One source of 
data about consumption can be found at:

http://bgp.potaroo.net/ipv4/


Considering that the amount of RIR allocated space these days that finds 
its way in the public routing table is close to 100% (see figure 2.4), and 
has been like this for the past 8 years or so, then the guiding metric is 
the growth of the amount of advertised space in the BGP table.

Taking the first order differential of this series of data shows that the 4 
year trend in address consumption is some 4 /8s per year, while the more 
recent data over the past 6 months reveals a consumption rate of up to 6 
/8s per year.

62 /8s at a rate of 6 /8s per year would be 10 years of consumption.

Usual caveats - this is, like any predictive exercise, replete with many 
assumptions about the future.

Taking a more conservative view of using the 10 /8s that are unallocated in 
the legacy B space, at current consumption rates this still looks like a 
pool of addresses that would provide some 20 months of allocations.  The 
analysis of the level of fragmentation of this space is also shown at the 
second URL - there are some 1100 /24 windows in this space, but also some 
350 /16 blocks and a smaller number of larger blocks (Figure 3.11)

Hope this helps,

regards,

    Geoff





At 08:46 AM 14/12/2004, Howard, W. Lee wrote:
>Please cite your sources.
>
>Since 0.1 years is less than a year, you must be suggesting that reclamation
>would save less than ten years, which would not be consistent with the
>references you almost give.
>
>Lee
>
> > -----Original Message-----
> > From: Randy Bush [mailto:randy at psg.com]
> > Sent: Monday, December 13, 2004 2:25 PM
> > To: Howard, W. Lee
> > Cc: ppml at arin.net
> > Subject: RE: [ppml] Reclamation (Was: Proposed Policy: PI
> > assignments for V6)
> >
> >
> > > It could be an interesting legal battle.  I haven't seen
> > contracts for
> > > assignments made before ARIN, so I can't say whether they
> > were made in
> > > perpetuity.  But I don't believe ARIN is required to
> > provide IN-ADDR
> > > delegations or WHOIS services for organizations with which
> > is has no
> > > relationship.
> >
> > actually, that committment was made to the fnc, nsf, and
> > others in order to get permission to form arin.
> >
> > > If ARIN reclaimed all unused address space, (since I'm sure nobody
> > > proposed reclaiming address space that's being used) how
> > much would we
> > > prolong the life of IPv4?  I suspect it would be less than a year.
> >
> > i suggest you're off by a decimal order of magnitude.  if the
> > book is too difficult, see george's and geoff's movie.
> >
> > randy
> >